Rack Power Calculator for PDU Capacity and A/B Failover Planning

How much power does a server rack use

There is no single useful answer for every rack. Equipment type, workload, configuration and power-management settings all affect consumption. A lightly populated networking rack and a rack of compute servers need different assumptions.

For an existing rack, record power during representative workload peaks. One reading taken during a quiet period is a poor basis for deciding whether another server will fit. For a planned deployment, use vendor configuration data, measurements from comparable equipment or an agreed design load.

Be careful with redundant power supplies. A server with two 1,200 W PSUs does not necessarily consume 2,400 W. Those labels describe power supply capacity, and may refer to DC output. The calculator’s equipment rows need the expected AC input power of the complete device.

When measuring an A/B rack, add simultaneous real-power readings from the two PDUs. If those readings already include all the servers, do not add the server-reported consumption again.

Calculate PDU capacity from voltage current and power factor

Watts describe real power. Volt-amperes describe apparent power. Power factor relates the two:

W = VA × power factor

At a power factor of 0.95, a 5 kW rack represents approximately 5.26 kVA. Lower power factor means more apparent power and current for the same real load.

The basic capacity equations are:

• Single phase: VA = voltage across the load × line current.
• Balanced three phase: VA = √3 × line-to-line voltage × current per line.

For a 400/230 V three-phase input, enter 400 V. For a single-phase load supplied at 208 V between two lines, select single phase and enter 208 V. Use the PDU’s input current rating per line, rather than adding outlet-bank ratings. These relationships are explained in the Schneider Electric Electrical Installation Guide.

The calculator uses the lower of the PDU input current rating and an entered upstream current limit. You can also enter a separate manufacturer real-power ceiling. This helps account for a constraint that a simple voltage-times-current calculation would miss.

Why two PDUs do not double redundant rack capacity

If a rack normally draws 3 kW from Feed A and 3 kW from Feed B, the surviving feed may need to supply the full 6 kW when the other feed is lost. Adding the two feed capacities would overstate the load that can survive that failure.

The calculator checks the normal split and both failure cases. Unequal feeds matter: a rack must remain within Feed A’s limits when B is unavailable, and within Feed B’s limits when A is unavailable. Vertiv’s guidance on redundant rack PDUs explains why surviving-feed loading is central to redundancy.

This assumes the equipment can operate from either feed. Single-corded devices need an appropriate transfer arrangement if they must survive the loss of their source. Independent upstream paths, available UPS capacity and transfer behavior also need verification.

Worked example for a 5 kW rack

The calculator starts with an illustrative rack using these values:

Each feed has approximately 10.81 kVA of apparent input capacity. At the entered load power factor, its nominal real-power capacity is about 10.27 kW. Applying the selected 80% planning ceiling gives 8.21 kW per feed.

The projected rack load is 5 × 1.20 = 6 kW. During normal operation, each feed carries 3 kW. After either feed fails, the survivor carries the full 6 kW, which remains within the modeled 8.21 kW planning ceiling.

Expansion headroom is approximately 2.21 kW beyond the entered growth allowance. If an additional server is expected to draw 500 W, the same 20% growth allowance makes its planning load 600 W. The calculator therefore shows room for three additional servers on electrical capacity alone. Rack space, cooling, connectors and outlet-bank limits may reduce that number.

Choose a planning ceiling that matches your installation

The calculator’s 80% starting value is an editable planning assumption. It is not a universal rule for every circuit, PDU or jurisdiction.

Read the manufacturer’s documentation carefully. Some ratings already describe usable continuous current under specified conditions. Applying a further planning percentage to that rating leaves additional reserve, which should be an intentional choice.

For identical A/B feeds with a 50/50 split and no extra failover allowance, an 80% ceiling on the surviving feed corresponds to 40% of nominal capacity on each feed during normal operation. Those percentages describe different operating conditions; they should not be treated as interchangeable limits.

Check individual phases as well as total rack power

A three-phase rack can appear to have spare capacity while one input line is carrying too much current. The main capacity model assumes balanced loading, so use the optional measured-current check to examine simultaneous L1, L2 and L3 readings from a PDU input.

The tool compares each line with the effective input current rating and your selected planning ceiling. It also reports the maximum current deviation from the three-line mean against an editable alert threshold.

This is a separate measurement snapshot. It does not scale the readings for growth or predict how current will redistribute after failover. Check branch and outlet-bank loading separately, and obtain representative post-failure readings through an approved test procedure where needed.

Turn rack power into a cooling estimate

Nearly all the electrical input to IT equipment ultimately becomes heat. The example’s projected 6 kW rack therefore represents approximately 6 kW of thermal load, or 20,473 BTU/h. The conversion is about 3,412 BTU/h per kW; Schneider Electric documents the watts-to-BTU/h relationship.

For liquid-cooled equipment, some heat is transferred directly into coolant. Total rack heat should therefore be separated into the portions handled by the liquid and air cooling systems. The calculator’s heat result does not establish that split.

The occupied-U input also gives projected kW per occupied rack unit. Treat this as a density indicator, rather than proof that airflow or cooling capacity is sufficient.

Use average load to estimate electricity costs

Capacity planning needs a load that represents demanding operating conditions. Electricity estimates need average power over time. Using peak rack power for every hour of the year can overstate consumption.

The calculator provides a separate average-load input. At 3.5 kW for 8,760 hours, annual rack electricity is 30,660 kWh. At USD 0.12/kWh, the modeled cost is USD 3,679.20.

Entering a facility PUE of 1.4 gives 42,924 kWh and USD 5,150.88 of facility-attributed electricity. This allocates infrastructure overhead to the rack using a facility ratio. It does not mean the rack produces 40% more heat.

These figures assume constant average load, tariff and PUE. They exclude demand charges, taxes and equipment costs. Use metered energy and appropriate facility data when assessing actual expenditure.

Compare the plan with AKCP monitoring data

The calculator gives you a planning baseline. Monitoring shows how closely the operating rack follows it.

Quicklime DCIM monitoring can bring compatible PDU and UPS data into view alongside AKCP environmental measurements. Supported SNMP and Modbus integrations provide access to measurements exposed by the connected equipment. Available points depend on the model, firmware and configuration.

Compare rack power trends with inlet temperatures and operating events. Review both feeds together, then investigate whether a capacity issue comes from rising total load, uneven distribution or a lower limit elsewhere in the power path. AKCP’s data center capacity planning and battery and UPS monitoring solutions provide more information on these monitoring options.

Before installing more equipment, check the projected rack load against normal and failure conditions, then verify the phase, branch, cooling and upstream constraints that the calculation cannot establish.